This is an automated email from the ASF dual-hosted git repository.
xiaoxiang pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/nuttx.git
commit 8e653753655a91829190c524e49aac3766336dc6
Author: ouyangxiangzhen <ouyangxiangz...@xiaomi.com>
AuthorDate: Fri Mar 28 17:15:34 2025 +0800
sched: Remove unnecessary tick++ for absolute wdog timer.
To improve the absolute timer accuracy, this commit move the tick++ to
relative wdog timer.
Signed-off-by: ouyangxiangzhen <ouyangxiangz...@xiaomi.com>
---
sched/signal/sig_timedwait.c | 10 +++++++++-
sched/timer/timer_settime.c | 11 ++++++++++-
sched/wdog/wd_start.c | 40 ++++++++++++++++++----------------------
3 files changed, 37 insertions(+), 24 deletions(-)
diff --git a/sched/signal/sig_timedwait.c b/sched/signal/sig_timedwait.c
index 219de4ee13..22cd6bd160 100644
--- a/sched/signal/sig_timedwait.c
+++ b/sched/signal/sig_timedwait.c
@@ -304,7 +304,15 @@ int nxsig_clockwait(int clockid, int flags,
if ((flags & TIMER_ABSTIME) == 0)
{
- expect = clock_systime_ticks() + clock_time2ticks(rqtp);
+ /* delay+1 is to prevent the insufficient sleep time if we are
+ * currently near the boundary to the next tick.
+ * | current_tick | current_tick + 1 | current_tick + 2 | .... |
+ * | ^ Here we get the current tick
+ * In this case we delay 1 tick, timer will be triggered at
+ * current_tick + 1, which is not enough for at least 1 tick.
+ */
+
+ expect = clock_systime_ticks() + clock_time2ticks(rqtp) + 1;
wd_start_abstick(&rtcb->waitdog, expect,
nxsig_timeout, (uintptr_t)rtcb);
}
diff --git a/sched/timer/timer_settime.c b/sched/timer/timer_settime.c
index 47e882ba21..9bc9578cb0 100644
--- a/sched/timer/timer_settime.c
+++ b/sched/timer/timer_settime.c
@@ -328,7 +328,16 @@ int timer_settime(timer_t timerid, int flags,
*/
delay = clock_time2ticks(&value->it_value);
- timer->pt_expected = clock_systime_ticks() + delay;
+
+ /* delay+1 is to prevent the insufficient sleep time if we are
+ * currently near the boundary to the next tick.
+ * | current_tick | current_tick + 1 | current_tick + 2 | .... |
+ * | ^ Here we get the current tick
+ * In this case we delay 1 tick, timer will be triggered at
+ * current_tick + 1, which is not enough for at least 1 tick.
+ */
+
+ timer->pt_expected = clock_systime_ticks() + delay + 1;
}
/* Then start the watchdog */
diff --git a/sched/wdog/wd_start.c b/sched/wdog/wd_start.c
index 3d84dd2d3e..4741ac0a41 100644
--- a/sched/wdog/wd_start.c
+++ b/sched/wdog/wd_start.c
@@ -119,14 +119,10 @@ static void wdentry_period(wdparm_t arg)
wdperiod->func(wdperiod->wdog.arg);
- /* Since we set `ticks++` at `wd_start_abstick`,
- * we need to use `expired - 1` here to avoid time drift.
- */
-
if (wdperiod->period != 0)
{
wd_start_abstick(&wdperiod->wdog,
- wdperiod->wdog.expired + wdperiod->period - 1,
+ wdperiod->wdog.expired + wdperiod->period,
wdentry_period, wdperiod->wdog.arg);
}
}
@@ -323,23 +319,6 @@ int wd_start_abstick(FAR struct wdog_s *wdog, clock_t
ticks,
return -EINVAL;
}
- /* Calculate ticks+1, forcing the delay into a range that we can handle.
- *
- * NOTE that one is added to the delay. This is correct and must not be
- * changed: The contract for the use wdog_start is that the wdog will
- * delay FOR AT LEAST as long as requested, but may delay longer due to
- * variety of factors. The wdog logic has no knowledge of the the phase
- * of the system timer when it is started: The next timer interrupt may
- * occur immediately or may be delayed for almost a full cycle. In order
- * to meet the contract requirement, the requested time is also always
- * incremented by one so that the delay is always at least as long as
- * requested.
- *
- * There is extensive documentation about this time issue elsewhere.
- */
-
- ticks++;
-
/* NOTE: There is a race condition here... the caller may receive
* the watchdog between the time that wd_start_abstick is called and
* the critical section is established.
@@ -437,6 +416,23 @@ int wd_start(FAR struct wdog_s *wdog, sclock_t delay,
return -EINVAL;
}
+ /* Calculate delay+1, forcing the delay into a range that we can handle.
+ *
+ * NOTE that one is added to the delay. This is correct and must not be
+ * changed: The contract for the use wdog_start is that the wdog will
+ * delay FOR AT LEAST as long as requested, but may delay longer due to
+ * variety of factors. The wdog logic has no knowledge of the the phase
+ * of the system timer when it is started: The next timer interrupt may
+ * occur immediately or may be delayed for almost a full cycle. In order
+ * to meet the contract requirement, the requested time is also always
+ * incremented by one so that the delay is always at least as long as
+ * requested.
+ *
+ * There is extensive documentation about this time issue elsewhere.
+ */
+
+ delay++;
+
return wd_start_abstick(wdog, clock_systime_ticks() + delay,
wdentry, arg);
}
